Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Tesla Bot Update – Sort and Stretch [video]

youtube.com

Post reply on HN
Print view View original

11–20 of 31 posts

Re: Tesla Bot Update – Sort and Stretch [video]

#11

Post by refulgentis »

Juicyy wrote: post #8
refulgentis wrote: post #7

Earlier quoted context omitted.

At 0:40 I see the sort task has magnets on the blocks. I'm curious, dropped out of state school economics: for robotics types is "sort blue and green" something you should be able to pull of as an undergrad, given equipment?

very easily. here is the lecture from my robotics course. Amazing professor. Dr. Sodemann was the best I had in my lifetime. https://robogrok.com/1-2-3_Cameras_and_Color.php Its actually easier than you would expect! you can use a process in image processing called "thresholding" where you basically filter the RGB or color values. Its very common and a little "hack" on compute. Computers "look" at images or video by…

Fascinating...my most recent deep work was around color so this really tickled me, ty!

Re: Tesla Bot Update – Sort and Stretch [video]

#12

Post by thejazzman »

elfbargpt wrote: post #5
7e wrote: post #2

Tesla can’t solve self driving, but it will produce a bot which can function in an environment which is a hundred times more open and chaotic than a road system? I don’t think so.

Honestly the stakes are a lot lower for a humanoid robot IMO. There's a lot of risk around releasing FSD cars on public roads as opposed to testing out robots in a factory / controlled environment. I'm actually really excited about the current state of robotics, and wouldn't be surprised if it ends up being the most impactful outcome of the most recent AI developments over the next 10-15 years. I know there's a lot o…

> risk around releasing FSD cars

This reads like they haven't already taken that risk with fatal consequences.

Not seeing how a robot, literally marketed as being powered by the FSD computer, could actually work unless the underlying FSD system actually works. Which it doesn't.

Re: Tesla Bot Update – Sort and Stretch [video]

#13

Post by somethoughts »

I've been generally curious about the benefits of the humanoid robot form factor.

I'll admit I don't know much about the state of the enterprise robotics/automation market but it would seem like the market would be limited by the fact that: 1.) The companies able to afford such robots would have much higher throughput capacity requirements and would want to setup much more customized automation. 2.) The companies in the sweet spot in terms of lower throughput capacity requirements would not be able to afford the upfront cost of such robots.

I think the primary use case would be perhaps if these could actually get in the sub $100K-$150K range end cost wise which seems a bit far off given the complexity I'm seeing here. Perhaps the idea is to go for the long game?

It feels like a version with just the top half of the humanoid would be more interesting if it could cut costs particularly since this thing likely needs to be tethered for reliable power and communications in a factory automation setting.

Re: Tesla Bot Update – Sort and Stretch [video]

#14

Post by cantaloupe »

sen wrote: post #4

Conveniently not labeled that the video is all at 4x speed. Using the “playback speed” menu to watch at 0.25x makes the human’s movements much more realistic and the robots movements much less impressive.

I think .25x is a bit slow, as .5x replay also looks reasonable, but you are right there is something like a 2.5–3.5x speed up. Makes the robot’s smooth motion much less impressive.

Re: Tesla Bot Update – Sort and Stretch [video]

#15

Post by elfbargpt »

thejazzman wrote: post #12
elfbargpt wrote: post #5

Earlier quoted context omitted.

Honestly the stakes are a lot lower for a humanoid robot IMO. There's a lot of risk around releasing FSD cars on public roads as opposed to testing out robots in a factory / controlled environment. I'm actually really excited about the current state of robotics, and wouldn't be surprised if it ends up being the most impactful outcome of the most recent AI developments over the next 10-15 years. I know there's a lot o…

> risk around releasing FSD cars This reads like they haven't already taken that risk with fatal consequences. Not seeing how a robot, literally marketed as being powered by the FSD computer, could actually work unless the underlying FSD system actually works. Which it doesn't.

I'm not sure about their marketing claims, but the robot in this video doesn't have wheels and isn't going 80mph on an interstate. Maybe the cars and this robot are using similar sensors, but I doubt there's a ton of overlap between the underlying system and the directions it gives. AGI doesn't exist yet

Re: Tesla Bot Update – Sort and Stretch [video]

#16

Post by panick21_ »

somethoughts wrote: post #13

I've been generally curious about the benefits of the humanoid robot form factor. I'll admit I don't know much about the state of the enterprise robotics/automation market but it would seem like the market would be limited by the fact that: 1.) The companies able to afford such robots would have much higher throughput capacity requirements and would want to setup much more customized automation. 2.) The companies in…

It seems a bit strange that Tesla can build multiple electrics cars for 150k$ but the robot that consists of much less material should cost that much.

The car has lots of compute power, likely they are sharing things with the car. The car has 5 cameras and a host of other sensors as well.

The car also requires the cost of a huge battery and tons of materials plus a lot of manual work for the interiors, plus cooling and so on.

Yes the robot has a lot more actuators and complex mechanical pieces but is that gone cost as much as 2 full cars?

What complexity do you think would be so costly here?

Re: Tesla Bot Update – Sort and Stretch [video]

#17

Post by SushiHippie »

cantaloupe wrote: post #14
sen wrote: post #4

Conveniently not labeled that the video is all at 4x speed. Using the “playback speed” menu to watch at 0.25x makes the human’s movements much more realistic and the robots movements much less impressive.

I think .25x is a bit slow, as .5x replay also looks reasonable, but you are right there is something like a 2.5–3.5x speed up. Makes the robot’s smooth motion much less impressive.

.5x makes the part where the human interrupts the robot while sorting look natural.

Re: Tesla Bot Update – Sort and Stretch [video]

#18

Post by mlindner »

panick21_ wrote: post #16
somethoughts wrote: post #13

I've been generally curious about the benefits of the humanoid robot form factor. I'll admit I don't know much about the state of the enterprise robotics/automation market but it would seem like the market would be limited by the fact that: 1.) The companies able to afford such robots would have much higher throughput capacity requirements and would want to setup much more customized automation. 2.) The companies in…

It seems a bit strange that Tesla can build multiple electrics cars for 150k$ but the robot that consists of much less material should cost that much. The car has lots of compute power, likely they are sharing things with the car. The car has 5 cameras and a host of other sensors as well. The car also requires the cost of a huge battery and tons of materials plus a lot of manual work for the interiors, plus cooling a…

> It seems a bit strange that Tesla can build multiple electrics cars for 150k$ but the robot that consists of much less material should cost that much.

Cars are only cheap because of massive economies of scale. This robot will probably be hand assembled, making it obviously quite expensive. Not to mention the need to recoup a lot the R&D investment money over relatively few units.

Re: Tesla Bot Update – Sort and Stretch [video]

#19

Post by mlindner »

elfbargpt wrote: post #5
7e wrote: post #2

Tesla can’t solve self driving, but it will produce a bot which can function in an environment which is a hundred times more open and chaotic than a road system? I don’t think so.

Honestly the stakes are a lot lower for a humanoid robot IMO. There's a lot of risk around releasing FSD cars on public roads as opposed to testing out robots in a factory / controlled environment. I'm actually really excited about the current state of robotics, and wouldn't be surprised if it ends up being the most impactful outcome of the most recent AI developments over the next 10-15 years. I know there's a lot o…

> he's just the money guy

This argument never makes sense as it's pure circular reasoning. He's had no money to invest in anything without taking from some of his companies (which is what happened to Twitter). You can't invest in Tesla with money from Tesla.

Re: Tesla Bot Update – Sort and Stretch [video]

#20

Post by migueloller »

somethoughts wrote: post #13

I've been generally curious about the benefits of the humanoid robot form factor. I'll admit I don't know much about the state of the enterprise robotics/automation market but it would seem like the market would be limited by the fact that: 1.) The companies able to afford such robots would have much higher throughput capacity requirements and would want to setup much more customized automation. 2.) The companies in…

I believe their target price is $20k. No idea if it costs that right now, I doubt it.
Post reply on HN